Skip to content

Complete proof of concept combining artificial intelligence, machine learning, and thousands of publicly available traffic cameras to place an entire city under surveillance with inference bots watching for a stolen car every single moment.

Notifications You must be signed in to change notification settings

cchighman/PriusWatchML

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

PriusWatchML

After my Prius was stolen in Seattle, I discovered there were thousands of publically available traffic cameras all over the state. I soon learned working with the city or agencies would prove futile as those cameras are on a tight leash by the ACLU. Ironically enough, while the local government's hands are tied to use them for any good purpose, any other person out there in the world are able to tap into a hidden gem of information and possibility with ease.

Over a 2-3 month window, object detection, CNN inference, and color heuristics were finely tuned to enable a basic Python script to run across many GPU-enabled Google Colab notebooks. These inference bots were assigned sections of the larger Seattle area and used a simple loop to constantly analyze traffic cameras images. After much iteration, the inference bots successfully detected my Prius C in downtown Seattle!

See the PDF "PriusWatchML - Experience.pdf" for an extremely detailed account of my experience, challenges, strategies, and how I finally made my way to success. There is also quite a bit of discussion on the topic of ethical AI as it was front and center during my tech talks sharing this across Microsoft AI communities. I held out publishing these details for nearly two years.

This proof of concept highlights many potential domains of possibility and concern, particularly as it relates to mining publically available traffic cameras over time in efforts to refine what could be one of the most disruptive datasets we've seen at this point in our AI evolution.

About

Complete proof of concept combining artificial intelligence, machine learning, and thousands of publicly available traffic cameras to place an entire city under surveillance with inference bots watching for a stolen car every single moment.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages